Heatwave Dataset - Metadata and Documentation

Background
A heatwave is broadly defined as a period of unusually hot weather lasing two to three or more
days. This hot weather is typically warmer than the historical averages for the affected region
and is also much warmer than the temperatures preceding and following the heatwave event.
Heatwaves have a direct impact on human health, the local environment, and the economy. Heat
stress from extreme heat events is the leading cause of weather-related deaths in the United
States. 1 A heatwave represents time when the ambient temperatures often outpace the body’s
ability to cool itself via perspiration, resulting in greater incidences of heat-related illnesses,
hospital admissions, and deaths among vulnerable populations. 2 Additionally, heatwaves can
result in crop failures, excessive load on the electricity grid from increased air conditioner use,
and can affect the health of the local ecosystem. 3
The average number and duration of heatwaves have been increasing across the U.S. in recent
decades 4 and this trend is expected to continue with climate change. Planning for emergency
response to future heatwaves will necessitate projections of how heatwaves may change in New
Jersey counties over the coming century. This dataset provides a measure of change in future
heatwave number and duration compared to a historical period of 1981–2010.

Defining a Heatwave
Heatwave Temperature Threshold
This dataset bases heatwave metrics on the daily average temperature. The average temperature
is a synthesis of the daily maximum and minimum temperatures. Daily maximums emphasize the
highest heat conditions a person may be exposed to, but daily minimums have been shown to be
more important for heat stress and illness when elevated for an extended period. The daily
average combines these effects so that neither is entirely discounted. Additionally, Anderson and
Bell (2009) 5 found that daily average temperature was the most consistent predictor of heatwave
mortality.
The exact temperature thresholds and durations used to identify heatwaves within temperature
records are not consistently defined in scientific literature. In this dataset, an extreme heat day is
defined as a day where the average temperature exceeds a predetermined threshold. We
calculated the threshold as the 95 th percentile 6–8 of 30 years (1981–2010) of daily average
temperature data during the warm season (May 1st–September 30 th). The benefit of using a
percentile approach is that each location can have a slightly different threshold. For example, the
climates of Capy May and Sussex Counties can be very different, Sussex being inland with more
topography (generally cooler) while Cape May is coastal, topographically flat, and further south
(generally warmer).
Heatwave Duration Threshold
Equally important in identifying a heatwave is the number of consecutive days that the
temperature threshold is exceeded. A heatwave is essentially multiple extreme heat days in
succession. We selected a minimum duration of three consecutive days of extreme temperatures
3,9,10 to define a heatwave. It is also common to select two days as a minimum as heat illness may

be present with a shorter exposure time. 4–6 While we elected to use a three-day minimum as it
represents a greater heat stress impact (exposure to heightened temperatures for a longer
duration), it is likely the calculated trends will hold true for shorter heatwaves as well.

Data Sources
The data in this analysis are composed of an observational dataset and a series of climate model
outputs. The observational data is a gridded dataset of about 20,000 NOAA Cooperative
Observer (COOP) stations across the conterminous United States. 11 We used NOAA’s Applied
Climate Information System (ACIS, http://www.rcc-acis.org/) to extract the daily average
temperature for the State of New Jersey for over 1981–2010 to establish a historical baseline of
annual heatwave statistics. These daily temperatures were averaged by county such that each
county had a single record of daily average temperatures from 1981–2010.
The modeled data were also retrieved through ACIS, again extracting daily average temperature
data by county. The modeled data was composed of 31 individual climate models that are part of
the Coupled Model Intercomparison Project 5 and were scaled regionally to New Jersey using
the Localized Constructed Analogues (LOCA) method. 12 The climate models provide data for
three separate time periods. The first was a historical period (1981–2010) to correspond to the
observational data. The other two were future time periods: 2036–2065 to represent the mid-21 st
century and 2070–2099 to represent the end of the 21 st century.
The projected future climates were modeled using 2 different greenhouse gas emissions
scenarios. The first, Representative Concentration Pathway (RCP) 4.5 represents moderate
greenhouse gas emissions though the end of the century. 13 The second, RCP 8.5 represents
unmitigated and growing greenhouse gas emissions by the end of the century 14 and is generally
considered the upper plausible limit of the magnitude of climate change. In general, under both
scenarios, the climate warms and heatwaves become stronger, but RCP 8.5 creates much higher
temperatures and greater heatwaves.
Table 1. Description of the Data Parameters used in this Analysis
Data Descriptions Description
Historic Time Period 1981–2010
2036–2065, defined as a 30-year interval centered on
Mid-Century Time Period
approximately 2050
2070–2099, as a 30-year interval centered on approximately
End-Century Time Period
2085
Representative Narrative descriptions about how global greenhouse gas
Concentration Pathways emissions (such as carbon dioxide and methane), air pollutants,
(RCPs) and land use may change over the 21st century.
The moderate greenhouse gas emissions scenario where
RCP 4.5 emissions peak in the mid-century and are reduced by the end
of the 21 st century.
The continued and expanding greenhouse gas emissions
RCP 8.5 scenario where emissions rise throughout the 21 st century
without any significant reductions.
A gridded dataset of daily average temperatures created by
Livneh and others (2013). 11 Data were extracted for New
Observed Data
Jersey counties for the time period of 1980–2010. Daily
temperatures were spatially averaged by county.
Gridded daily average temperature datasets of 31 climate
models available through ACIS. Data were extracted for New
Jersey counties for historic, mid-century, and end-century time
Modeled Data
periods. Daily temperatures were spatially averaged by county
and calculated heatwave metrics of each model were combined
using a weighted average. 12

Heatwave Calculations and Change

Heatwaves were calculated for the observed dataset and for each model output for the historical,
mid-century, and end-century time periods. First, the 95 th percentile threshold was calculated for
each county for the historical period using both the observed and modeled data. For each time
period, the annual average number, the average annual duration, and the maximum average
duration (the average of the longest heatwave each year over 30 years of data) of the identified
heatwaves were calculated with respect to the 95 th percentile threshold.
The modeled heatwave statistics were combined into a muti-model average using a weighted
averaged defined by the LOCA downscaling approach. 12 This process was utilized to produce
averages for the historical, mid-century, and end-century time periods for both RCP 4.5 and RCP
8.5. To project the statistics of future heatwaves, it is inadvisable to just use the model outputs.
Instead, a common approach is to calculate the percent change between the modeled historical
period and the future time periods. These percent change values were applied to the observed
historical heatwave data to yield future heatwave projections more firmly anchored to real
values. This approach is conceptually visualized in Figure 1. These projections are not perfect
but are likely to be more accurate when using this approach.

Figure 1. Flowchart of how modeled heatwave metrics are averaged and how a percent change
between modeled historical and modeled projected averages is applied to observed heatwave
metrics. This example used a future period of 2070–2099, but the approach is the same for the
mid-century time period and for both RCP 4.5 and 8.5.

Finally, a “percentile range” of future heatwave change was defined for each future scenario. The
range is described in the heatwave data file by the 10th and 90 th percentile columns for each
parameter (heatwave number, duration, and maximum duration) for the future scenarios. The
percentiles are computed by taking the heatwave outputs for the 31 climate modes before
averaging and removing the lower 10 th percentile and upper 90 th percentile of individual outputs
from the 31 climate models. This approach provides a constrained range of model projections
akin to (but different from) a confidence interval by excluding the highest and lowest 10% of the
individual model outputs. For example, the annual number of heatwave events is computed for
each model, so a single county will have 31 separate modeled values for the number of
heatwaves for a future time period and emissions scenario (such as mid-century under RCP 4.5).
The 10 th and 90 th percentiles for the number of heatwaves for this period of are calculated from
these 31 data values. This process was repeated for the other time periods, emissions scenarios,
and counties. The percentile values where then computed as a percent change from the modeled
historical mean and the percent change applied to the observed number of heatwaves to define
the future limits relative to the observed values.

Table 2. Descriptions of the Heatwave Data Fields Presented in the Heatwave_Data.xlsx File
Heatwave Data Field Description
The threshold above which an average daily temperature
constitutes an extreme heat day. Three or more consecutive
days above this threshold constitutes a heatwave. The
95 th Percentile Temperature
threshold is different for each county to account for local
(F)
effects. The threshold is calculated as the 95 th percentile of
observed daily average temperatures during the warm season
for 1981–2010.
The annual average number of heatwaves a county
Mean Annual Heatwave
experiences/is projected to experience based on 30 years of
Number
data.
The annual average duration of the heatwaves a county
Mean Annual Heatwave
experiences/is projected to experience based on 30 years of
Duration (Days)
data.
The average annual maximum duration of the heatwaves a
Max Annual Heatwave
county experiences/is projected to experience based on 30
Duration (Days)
years of data.
The 10 percentile value for a given parameters (heatwave
th
10% Annual Heatwave
number, duration, or maximum duration) based on the spread
Number/Heatwave
of results of 31 climate models. Only available for projected
Duration/Max Heatwave
future heatwaves. Consider this as the lower end of a
Duration
“constrained range” of change from observed conditions.
The 90 th percentile value for a given parameters (heatwave
90% Annual Heatwave
number, duration, or maximum duration) based on the spread
Number/Heatwave
of results of 31 climate models. Only available for projected
Duration/Max Heatwave
future heatwaves. Consider this as the upper end of a “likely
Duration
constrained” of change from observed conditions.

Interpretation
The future heatwave data within the Heatwave_Data.xlsx file are categorized by RCP emissions
scenario, RCP 4.5 or RCP 8.5, and by the future time period. These data are annual averages, not
the total for the time period, so many of the values are not whole numbers. Additionally, since
these values are averages for 30-year periods of time, one should not try to pinpoint an exact year
when these conditions are most likely to occur but should use them as a guide for the general
mid- and end-century time periods. Following this approach, these projections are a guide, and
the absolute values presented should be viewed as an approximation of projected change, not the
exact future conditions. The 10 th and 90 th percentiles provide a constrained range of possible
change from the observed conditions and can be interpreted as 80% of the modeled outputs agree
that the future heatwave metrics are within the range bounded by the 10 th and 90 th percentiles.
Finally, the future heatwave metrics are relative to the threshold defined for the historical period.
A new threshold is not determined for each time period. The reason for this approach is to help
clarify the magnitude of future changes relative to the observed period’s extreme conditions.
Under a changing climate, what could be defined as an extreme temperature day may change as
people acclimate and emergency infrastructure adapts to warmer temperatures. However, it is
most useful to present these changes relative to conditions that managers and the public have
experienced.

References
1. NOAA. Weather Related Fatality and Injury Statistics. National Oceanic and Atmospheric
Administration (NOAA).
2. Moran, D. et al. New Jersey Climate and Health Profile Report. New Jersey Climate
Adaptation Alliance. (2017).
3. Agel, L., Barlow, M., Skinner, C., Colby, F. & Cohen, J. Four distinct Northeast US
heatwave circulation patterns and associated mechanisms, trends, and electric usage. Npj
Clim. Atmospheric Sci. 4, 1–11 (2021).
4. Habeeb, D., Vargo, J. & Stone, B. Rising heatwave trends in large US cities. Nat. Hazards
76, 1651–1665 (2015).
5. Anderson, B. G. & Bell, M. L. Weather-Related Mortality. Epidemiol. Camb. Mass 20, 205–
213 (2009).
6. Anderson, G. B. & Bell, M. L. Heatwaves in the United States: Mortality Risk during
Heatwaves and Effect Modification by Heatwave Characteristics in 43 U.S. Communities.
Environ. Health Perspect. 119, 210–218 (2011).
7. Nairn, J. & Fawcett, R. Defining heatwaves: heatwave defined as a heat impact event
servicing all community and business sectors in Australia. The Centre for Australian
Weather and Climate Research, CAWCR technical report 060, 84 (2013).
8. Argüeso, D., Di Luca, A., Perkins-Kirkpatrick, S. E. & Evans, J. P. Seasonal mean
temperature changes control future heatwaves. Geophys. Res. Lett. 43, 7653–7660 (2016).
9. Meehl, G. A. & Tebaldi, C. More Intense, More Frequent, and Longer Lasting Heatwaves in
the 21st Century. Science 305, 994–997 (2004).
10. Peng, R. D. et al. Toward a Quantitative Estimate of Future Heatwave Mortality under
Global Climate Change. Environ. Health Perspect. 119, 701–706 (2011).
11. Livneh, B. et al. A Long-Term Hydrologically Based Dataset of Land Surface Fluxes and
States for the Conterminous United States: Update and Extensions. J. Clim. 26, 9384–9392
(2013).
12. Pierce, D. W., Cayan, D. R. & Thrasher, B. L. Statistical Downscaling Using Localized
Constructed Analogs (LOCA). J. Hydrometeorol. 15, 2558–2585 (2014).
13. Thomson, A. M. et al. RCP4.5: a pathway for stabilization of radiative forcing by 2100.
Clim. Change 109, 77–94 (2011).
14. Riahi, K. et al. RCP 8.5—A scenario of comparatively high greenhouse gas emissions. Clim.
Change 109, 33 (2011).